摘要
本文在研究了ART算法、Field Theory和ARTMAP等算法的基础上提出了FTART算法,它将自适应谐振理论、域理论等的优点有机结合.以样本在实例空间中出现的概率为启发信息修改网络中的分类,并采用了不同于其它算法的解决样本间的冲突和动态扩大分类区域的方法.测试结果表明,FTART算法具有比较好的性能.
It is presented that neural network learning algorithm FTART which is based on our research of several algorithms such as ART, Field Theory, ARTMAP, etc. and combines the advantages of these algorithms. FTART makes use of the occurrence probability of examples in the instance space as heuristic information to modify the classification and employs a different method for resolving the conflicts among examples and dynamically expanding classification area. Some benchmark test results show that FTART has a better performance.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1996年第4期330-336,共7页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金
